If you haven’t already, check out Arun Ulag’s hero blog “Microsoft Build 2026: Building Agentic Apps with Microsoft Fabric and Microsoft Databases” for a complete look at all of our Microsoft Build announcements across our Fabric and database offerings.

___________________________________

Pipelines in Microsoft Fabric Data Factory have always been successful at moving and transforming data. With this set of Build 2026 releases, they take on a much bigger job: orchestrating the people and processes around your data, not just the data itself. Alongside that, we’re shipping a list of authoring, runtime, and Airflow improvements that make pipelines and DAGs easy to build, easy to run reliably, and easy to promote across environments.

Human-in-the-loop with the new Approval activity (Preview)

The new Approval activity brings human decision-making directly into pipelines. Configure an email approval request, and pipeline execution pauses until the approver responds — then continues, branches, or fails based on the decision. No external systems to stitch together, no manual coordination steps.

This unlocks scenarios pipelines couldn’t cleanly handle before:

• Business sign-off before promoting data or a model to production
• Validation checkpoints for sensitive operations
• Exception-handling workflows that need a human in the loop

To learn more, refer to the Pipelines are evolving beyond ETL blog post.

Figure: The approval activity’s approver experience.

Materialized Lake View and SQL refresh updates

Refresh Materialized Lake View activity (Preview) triggers refreshes of your Materialized Lake View from inside a pipeline. Fabric decides automatically whether a full refresh, incremental refresh, or no refresh is needed based on what changed upstream so you don’t burn compute on refreshes you don’t need.

Learn more from the Materialized Lake View Refresh Activity documentation.

Refresh SQL Endpoint activity (Generally Available) does the same for the SQL analytics endpoint, so reporting and analytics consistently reflect the latest data as part of your orchestration logic.

Learn more from the Refresh SQL Endpoint activity documentation.

Figure: Refresh Materialized Lake View activity settings.

Conditional activity retries (Preview)

When a pipeline activity fails, you shouldn’t have to choose between retrying everything or nothing. Conditional activity retries let you define exactly when an activity should retry based on the error it returns — error code, error message, or both, with AND/OR logic. Your pipelines recover automatically from transient failures and fail fast on errors that won’t self-resolve. No wrapper activities, no custom error-handling code, no wasted compute.

This is especially useful for long-running pipelines that call external APIs, run Databricks jobs, or execute stored procedures where some error classes are retriable and others aren’t.

Learn more from the conditional retry pipeline activity documentation.

Figure: Use Conditional Retries to build workflows that retry based on conditions.

A modern pipeline canvas (Preview)

For pipelines with dozens or hundreds of activities, the canvas matters as much as the activities do. The new modern Pipeline node experience improves zoom behavior, gives you a cleaner layout for complex orchestration logic, and makes the canvas more responsive at scale.

Figure: New pipeline UI node updates at the zoom-out level.

Learn more from the pipeline canvas documentation.

Key enhancements in Variable Libraries for connections, items, and Airflow

Connection and item references in pipelines (Preview). Variable Libraries now support two new variable types: connection references to securely point at external sources without embedding connection details, and item references to dynamically link to Lakehouses, notebooks, or other pipelines. Promote the same pipeline definition across dev, test, and prod — connections and resources resolve based on the active value set.

Learn more from the Variable Library in pipelines documentation.

Variable Library support in Apache Airflow job (Preview). The same workspace-level Variable Library you use for pipelines now drives DAGs in Fabric’s managed Airflow. Define one library with multiple value sets, mark one active per workspace, and Airflow resolves the right values at runtime. No DAG edits, no redeployment as you promote.

Learn more from the Fabric Apache Airflow job documentation.

Figure: Now you can use Variable Libraries in your Fabric Apache Airflow jobs

Migration made easy from ADF

We’ve made migrating to Fabric Data Factory even easier from Azure Data Factory (ADF). Now you can find the migrate button right in your Fabric workspace under the workspace Migrate menu. Clicking the migration button is fast and has no cost. You will instantly gain access to an ADF assessment report evaluating your Data Factory’s readiness to migrate to Fabric. Then you can make the decision when you’re ready to click the final migration button to complete the process.

Figure: Click on Migrate from your Fabric workspace view to find the Data Factory migration tile and start your migration from ADF today.

Wrapping up

Combined, these Build 2026 releases reshape the capabilities of Fabric Data Factory pipeline.

Approval activities and conditional retries bring human decision-making and resilience directly into the flow. Materialized Lake View and SQL endpoint refresh activities keep downstream analytics in lockstep with your orchestration.

The modern canvas makes complex pipelines easier to author and navigate, while expanded Variable Libraries now cover connections, items, and Airflow, making promotion across development, test, and production genuinely seamless. With in-workspace ADF migration lowering the barrier to entry, getting started is faster than ever.

Pipelines are no longer just moving data; they’re orchestrating the people, processes, and environments around them.

Try these capabilities in your Fabric workspace—and tell us what’s working, what’s not, and what you want to see next using the Microsoft Fabric Ideas site.